Assessment - Generating Welfare Indices
There is a general consensus that no single parameter can be used to accurately assess a fish’s welfare (Huntingford et al. 2006). However, there are challenges associated with measuring numerous indicators in regard to the most appropriate means of integrating and interpreting them. This is especially true in the case of deciding on the relative level of importance that should be attributed to each indicator. The following have been suggested as methods in which parameters can be integrated for on-farm welfare assessment for terrestrial animals (Spoolder et al. 2003):
- Scoring Systems: usually involve a threshold limit e.g. pass/fail for a number of parameters that are selected and weighted by experts.
- Decision Support Systems: Quantify relevant information from scientific literature and link to welfare needs which are assigned relative weightings e.g. need for shelter, food, to express natural behaviour etc.
- Multivariate statistics to determine relative weights for a number of parameters selected by experts.
- Post-hoc interpretation of commonly applied parameters (selected from the literature) to identify relevant (and irrelevant) parameters and allocate a weighting to each.
- Qualitative assessment – integrating parameters through ‘whole animal’ observations with allocations of scores
A technique recently applied in several studies attempting to assess fish welfare is principal components analysis (PCA). PCA is a multivariate statistical technique that can be used as a data reduction tool to produce principal components (PCs) that are based on observed coherence between numerous simultaneously measured parameters (Turnbull et al. 2004; Vamaros et al. 2006; North et al. 2006). The factor scores for valid PCs can be incorporated into statistical models as dependent variables in the same way as individual parameters would be. This technique appears to be robust and the indices generated appear to reflect biologically meaningful relationships between different parameters. PCA serves as a useful tool.
Advantages of using this type of analysis include:
- The removal of the subjectivity associated with trying to apply relative weightings to different parameters,
- Reduction of the chance of a Type I statistical error (i.e. falsely observing a statistical difference), which is greatly increased by carrying out numerous univariate analyses of individual parameters
- Models using PC's as dependent variables generally account for a much greater degree of observed variability than individual parameters
There are however limitations of using PCA to generate welfare indices. There is a degree of subjectivity associated with interpreting the generated PCs and the observed associations observed can sometimes be counterintuitive. The PCs generated are merely a reflection of statistical coherence between parameters and their usefulness will largely be dependent on the initial selection of appropriate parameters.